Heterogeneous Expression of Nuclear Encoded Mitochondrial Genes Distinguishes Inhibitory and Excitatory Neurons

Abstract

Mitochondrial composition varies by organ and their constituent cell types. This mitochondrial diversity likely determines variations in mitochondrial function. However, the heterogeneity of mitochondria in the brain remains underexplored despite the large diversity of cell types in neuronal tissue. Here, we used molecular systems biology tools to address whether mitochondrial composition varies by brain region and neuronal cell type in mice. We reasoned that proteomics and transcriptomics of microdissected brain regions combined with analysis of single-cell mRNA sequencing (scRNAseq) could reveal the extent of mitochondrial compositional diversity. We selected nuclear encoded gene products forming complexes of fixed stoichiometry, such as the respiratory chain complexes and the mitochondrial ribosome, as well as molecules likely to perform their function as monomers, such as the family of SLC25 transporters. We found that the proteome encompassing these nuclear-encoded mitochondrial genes and obtained from microdissected brain tissue segregated the hippocampus, striatum, and cortex from each other. Nuclear-encoded mitochondrial transcripts could only segregate cell types and brain regions when the analysis was performed at the single-cell level. In fact, single-cell mitochondrial transcriptomes were able to distinguish glutamatergic and distinct types of GABAergic neurons from one another. Within these cell categories, unique SLC25A transporters were able to identify distinct cell subpopulations. Our results demonstrate heterogeneous mitochondrial composition across brain regions and cell types. We postulate that mitochondrial heterogeneity influences regional and cell type-specific mechanisms in health and disease.

Keywords: GABA; glutamate; mitochondria; mitochondrial ribosome; respiratory chain; solute transporter.

Figure 1.

RNAseq analysis of microdissected mouse brain regions. A₁, A₂, Volcano plots of cortex compared with hippocampus, cortex compared with striatum, and hippocampus compared with striatum from adult male mice (n = 5). Threshold for significance was set at p < 10⁻³ and log₂ fold change at 1. Color code symbols depict the fold of change below or above the thresholds. A₁, All transcripts quantified using DSeq2 annotated to the mouse genome GRCm38. A₂, All nuclear transcripts encoding subunits of the respiratory chain complexes, the mitochondrial ribosome, and the SLC25A family of transporters. Note that scarce numbers of these nuclear encoded mitochondrial transcripts show modest expression differences among brain regions. B, Validation of the RNAseq results using as a comparison the in situ hybridization data from the Allen Mouse Brain Atlas. The 100 most upregulated and downregulated genes when comparing cortex and hippocampus by RNAseq were correlated with the differences reported by the Allen data. C, t-SNE analysis of the RNAseq data presented in A₁. D, t-SNE analysis of the data presented in A₂.

Figure 2.

TMT proteomic analysis of microdissected mouse brain regions. A, Multiscatter plots with all individual biological replicates used for TMT quantifications. Insets, Pearson similarity coefficients and PCA of samples in multiscatter plots. B, C, Volcano plots of cortex compared with hippocampus, cortex compared with striatum, and hippocampus compared with striatum from adult male mice (n = 5). Threshold for significance were set at p < 10⁻³ and log₂ fold change at 1. Color code symbols depict the fold of change below or above the thresholds. B, All proteins quantified in brain samples with inset Venn diagrams depicting the overlap between our TMT data (blue) and label-free quantifications by Sharma et al. (2015; pink). Representation factor and p values were estimated with an exact hypergeometric probability test. C, All nuclear encoded subunits of the respiratory chain complexes, the mitochondrial ribosome, and the SLC25A family of transporters. Note the abundant nuclear encoded mitochondrial proteins differentiating brain regions. D, Heat maps of the proteins that show the most pronounced changes based on the q value and magnitude of the difference. E, t-SNE analysis of the proteome data presented in B. F, t-SNE analysis of the data presented in C. Note that the best clustering of brain regions is obtained with the nuclear encoded mitochondrial proteins described in C. G, Simple linear correlation analysis of expression differences across brain regions. Proteins belonging to respiratory chain complexes, the mitochondrial ribosome, and the SLC25A family of transporters are color coded. Note the differences in slopes. p values describe the differences between adjacent correlation plots slopes obtained with Prism. Shaded area represents the 95% confidence interval. See Extended Data Figure 2-1 for list of protein hits with p < 10⁻³ and log₂ fold change of least 1.

Figure 3.

Nuclear encoded mitochondrial transcripts differentiate neurons by neurotransmitter identity and anatomical location. A, Volcano plots were assembled using the Allen single-cell RNAseq dataset. A total of 50,002 pyramidal glutamatergic neurons were compared with 22,745 GABAergic interneurons. Volcano plots are organized by subunits belonging to the mitochondrial ribosome, electron chain complexes I to V, and the SLC25A family of solute transporters. The mitochondrial ribosome and the SLC25A family of transporters are the most dissimilarly expressed transcripts when comparing GABAergic with glutamatergic neurons. B₁, t-SNE cell atlas generated with the expression levels of all transcripts encoding mitochondrial ribosome subunits. The t-SNE atlas encompasses >20 areas of mouse cortex and hippocampus, totaling 76,307 cells. Color codes denote brain regions annotated by the Allen Brain Atlas. B₂ shows B₁ data after 100 consecutive permutations. Anatomical segregation is lost. C, Diagram explaining strategy for cell type and anatomic callout in t-SNE atlases. GABAergic neurons were color-coded green and glutamatergic neurons were color coded gray. Cell type and anatomic region were marked by a triangle. D, t-SNE atlas shown in B₁ that was layered with the neurotransmitter identity of cells and anatomic location of cells (triangles). GABA, parvalbumin (Pval), somatostatin (Sst), γ-synuclein (Sncg), vasointestinal peptide (Vip), and lysosomal-associated membrane protein family member 5 (Lamp5) denote markers defining specific interneuron subpopulations. E, t-SNE atlas shown in B₁ but layered with the subtype of interneuron (triangles).

Figure 5.

Differential expression of selected nuclear encoded mitochondrial transcripts further differentiates neuronal subpopulations. A–C, t-SNE cell atlases built with the subunits of the mitochondrial ribosome (A), the SLC25A family of transporters (B), and the electron transport chain Complex I (C). t-SNE cell atlases were overlaid with heat maps of the expression levels of selected subunits of protein complexes or transporters. B₁, Transporters diffusely expressed across brain regions or showing specific patterns of expression. B₂, Transporters Slc25a22, Slc25a37, and Slc25a42 preferentially expressed in glutamatergic cells (see Fig. 3A). B₃, SLC25A transporters annotated in the SFARI database associated with autism spectrum disorder (SLC25A12, SLC25A27, and SLC25A39) or whose expression is altered in postmortem autism brain samples (SLC25A12, SLC25A14, and SLC25A27).

Figure 4.

Families of nuclear encoded mitochondrial transcripts differentiate neurons by neurotransmitter identity and anatomical location. A, B, t-SNE cell atlases were generated with the expression levels of nuclear transcripts either encoding subunits of the respiratory chain complexes I to V, the mitochondrial ribosome, or the SLC25A family of transporters. A, t-SNE atlases encompasses >20 areas of mouse cortex and hippocampus, totaling 76,307 cells in each case. Color codes denote brain regions annotated by the Allen Brain Atlas. B, GABAergic neurons color-coded green and glutamatergic neurons color coded gray. PV-positive and somatostatin-positive cells were marked by a triangle. Note that families of transcripts can segregate cells by their lineage and anatomic origin.